AI Strategies: Learning from the last ‘gold rush’ to cloud

With the hype and excitement about AI unlikely to subside anytime soon, organisations are going to have to stay laser-focused on how to harness its capabilities effectively and not jump headfirst onto the AI bandwagon. But alternatively, they mustn’t be overly cautious, as those that take too long to leverage its benefits run the risk of losing out to faster, more agile competitors.

It brings to mind lessons that should have been learned after the initial rush to adopt cloud technologies. Driven by the fear of missing out and anticipating quick returns, organisations of all shapes and sizes raced to move their applications to the cloud. However, many had poorly conceived strategies which led to a catalogue of issues, most of which could have been averted, or at least mitigated, with better planning.

Lessons from the cloud

At the time, organisations lacked an understanding of the specific requirements needed for different types of data. Selecting the right cloud service model was frequently mishandled as businesses didn’t fully appreciate how their needs would be best served by the different options available.

Often, operations teams underestimated the complexity of the migration process, resulting in much longer migration timeframes and reduced solution quality. They also failed to recognise the importance of scalability giving rise to performance issues and business case challenge when usage costs started to exceed the parameters for savings.

Security teams relied on inadequate risk assessments which led to challenges in data security and compliance, particularly relating to where data resided and how it was protected. Add in insufficient access controls, and it all amounted to some very disappointing outcomes and wasted resources.

These mistakes highlighted the need for better advance planning with comprehensive strategies, including security measures and staff training, to ensure successful and secure adoption of new technologies. Those organisations that enjoyed outstanding cloud implementations had put together transformation strategies aligned to goals, with carefully planned, systematic roll-outs. Fundamentally, they recognised that simply moving to cloud technology itself didn’t guarantee positive business outcomes – which will be the same for AI.

Start as you mean to go on

Before diving headfirst into the world of AI, business leaders need to define what they are trying to achieve, whether that’s growth, efficiency, profitability, or a combination of factors. Then, a plan to

introduce AI can be organised around clear business objectives and milestones to deliver the desired goals.

Currently, there’s a lack of AI-specific implementation best practices to follow and experienced professionals who have delivered such projects are hard to find. Additionally, there’s a large dose of AI-washing happening with data analytics offerings simply being repackaged under the AI banner. Therefore, decision-makers need to be wary of unfounded marketing and delivery claims before partnering with vendors and consultancies.

The importance of data strategy

It’s hard not to get carried away with the magnitude of AI, but it can only be as smart as the information that’s put into the tools and platforms it serves. For the less tech-savvy, the term AI can be misleading implying that its intelligence will correct mistakes and inaccuracies in source data. But bad data will deliver erroneous, and potentially, damaging results.

Users also need to be careful when instructing language learning models otherwise results may be skewed. Verification of analysis and reports generated by AI should always be incorporated into a review process to ensure that major decisions are not based on inaccurate responses.

It’s equally important to ensure that the models are working from all relevant data sets. Organisations should be aware that the adoption of cloud technology may have made AI initiatives harder to handle if data has become fragmented. While cloud services have enabled seamless collaboration and data access across departments and locations, large data silos have been created too as different functions have chosen different platforms. Tying together big data strategies around disparate data sets is complex, and organisations should not leap straight into AI without a well-defined data management strategy in place.

Don’t overlook security

The lack of maturity concerning AI security represents a serious risk that mustn’t be underestimated. There are already instances of accidental misuse such as the data leak incident at Samsung, along with warnings of how malicious actors could manipulate applications, for example using chatbot prompt injections as outlined by the NCSC recently. Making security and compliance an integral part of AI planning from the start is vital, as well as including security training as part of implementation schedules. Otherwise, starting out on the wrong foot will inevitably lead to problems further down the line and increase the risk of serious data breaches.

Learning from new AI challenges

It would be naive to think that mistakes won’t be made along the AI road, but taking stock of past experiences, like cloud projects, will help inform today’s planning. No one knows exactly how AI will

develop in the future, but it is already changing the way people live and work. Successfully leveraging AI depends on starting out with well thought out strategies with clear objectives. This will help organisations stay focused on desired outcomes while retaining the capacity to deal with and learn from the new challenges that will continue to arise and evolve throughout the AI journey.